But AI is only called AI as long as it is "magical", as soon as it gets useful and is integrated in to real products we start calling it something else, like machine translation.

@jwz posted a very good description of AI:
jwz.org/blog/2023/03/ai-is-mag…

@aha and what people consider AI now are just large language models. There is still no intelligence involved. They don't think, they don't know logic, they just do completion based on the huge dataset they were trained on. The term AI has been used for all these things, and is still being used for how we translate and render speech now.

AI does mean something, just not only whatever is in fashion today. The term is half a century old and has been used throughout for lots of things. It's more than generative models.

@Gjoel Most of the things are machine learning (ML), not AI. It is absolutely crucial not to call ML AI, because the term AI causes people to assign it magic properties which it does not have.

Nobody who taught me statistics, ML, or the very basics of neural networks (we didn't have DNNs yet back then) called this stuff "AI". But yes, my sample size isn't that large (three universities, lots of books).

@aha In that case you shouldn't call anything AI? AI is a bit of a marketing term, so it's fuzzy by definition. What I have seen, people I work with, people in the industry use AI for things where you train a computer on data to do something with the resulting model (broadly speaking).

What definition of AI would you use?

@Gjoel I believe it is important to call a spade a spade. Otherwise, people get hurt.

I'm willing to call something AI the moment it exhibits at least some properties that we associate with intelligence, like reasoning or abstraction. As far as I know, we haven't reached that level yet and I assume it will take us some time, especially because we still haven't figured out how brains actually work. But I'm no expert in AI or brain research.